Dashboard

Overview

The attune-rag dashboard is a living-docs view of a registered corpus that moves data through three stages — snapshot, render, and display — to give you both a shareable HTML report and an at-a-glance terminal summary.

Three-stage pipeline

Each stage is independent and has its own entry point, so you can run them individually or chain them together.

1. Snapshot (refresh) build_snapshot(corpus_package) benchmarks the corpus against your queries.yaml file and returns a plain Python dict. If queries.yaml is missing, the function still returns a partial snapshot that includes an error field rather than raising an exception. This dict is the single source of truth that the other two stages consume.

2. Render (render) render(out, snapshot, title) takes the snapshot dict, serializes it as JSON, and injects it into an HTML template by replacing the sentinel strings __ATTUNE_SNAPSHOT__ and __ATTUNE_TITLE__. The result is a self-contained HTML file written to out — no external data files required to open it in a browser.

3. Display (show) display(snapshot, console) pretty-prints the same snapshot dict to the terminal using Rich. Pass your own Console instance to control output destination, or omit it to use the default.

How the pieces fit together

queries.yaml
     │
     ▼
build_snapshot()  ──►  snapshot dict  ──►  render()   ──►  report.html
                                      │
                                      └──►  display()  ──►  terminal

The snapshot dict is the interface between stages. Because render() and display() both accept an arbitrary dict, you can supply a snapshot from any source — not only from build_snapshot().

Entry points

When the dashboard matters

Use the dashboard when you need to verify that a corpus is answering its benchmark queries correctly, share a point-in-time report with teammates (use the rendered HTML), or spot freshness regressions quickly without leaving the terminal (use display).